Microsoft study names 20 jobs least likely to be impacted by AI — healthcare and blue‑collar roles top the list
A Microsoft Research analysis mapped 200,000 anonymized Bing Copilot interactions to U.S. job tasks and found manual, blue‑collar work and hands‑on healthcare support roles have the lowest AI applicability scores — meaning they are least suited to today’s LLMs.
- Key takeaways:
- Microsoft created an “AI applicability score” by mapping Bing Copilot conversations to O*NET/BLS tasks.
- Jobs requiring physical presence, dexterity or face‑to‑face care score lowest and are currently least well supported by LLMs.
- The score measures where LLMs can assist today; it does not equal job loss and excludes robotics/vision automation.
- Local implications: trades and hands‑on healthcare professions may warrant prioritized training and wage policy.
How Microsoft measured AI’s reach into jobs
Microsoft’s team analyzed 200,000 anonymized, privacy‑scrubbed conversations with Bing Copilot from 2024 and mapped each interaction to work activities and U.S. occupations (O*NET/BLS taxonomy). The result was an “AI applicability score” estimating how much of a job’s daily tasks could be performed or strongly assisted by today’s large language models (LLMs). High scores point to work built around writing, research, communication and routine digital tasks; low scores point to jobs that need hands‑on work, machines, or in‑person care. See the Full study and the technical paper for methodology and data.
The 20 jobs Microsoft finds least impacted by generative AI
Microsoft’s rankings, summarized across research materials and reporting, cluster occupations with the lowest AI applicability scores. These roles are currently least likely to be directly affected by LLMs — though other forms of automation could matter later.
- Roofers — outdoor, physically demanding roofing tasks.
- Helpers—Roofers — on‑site manual assistance for roofers.
- Tire Builders — hands‑on, machine‑linked tire production.
- Gas Compressor and Pumping Station Operators — on‑site control of industrial systems.
- Surgical Assistants — direct physical support in operating rooms.
- Massage Therapists — hands‑on therapeutic care requiring touch and dexterity.
- Ophthalmic Medical Technicians — in‑person vision testing and equipment use.
- Industrial Truck and Tractor Operators — forklift and yard vehicle operation.
- First‑Line Supervisors of Firefighting and Prevention Workers — on‑scene leadership in emergencies.
- Cement Masons and Concrete Finishers — physical, site‑specific concrete work.
- Dishwashers — manual kitchen work and equipment operation.
- Machine Feeders and Offbearers — loading and unloading production machines.
- Packaging and Filling Machine Operators and Tenders — tending industrial packaging lines.
- Medical Equipment Preparers — sterilizing and preparing tools in care settings.
- Home Health Aides / Personal Care Aides — hands‑on care for patients at home or in facilities.
- Nursing Assistants / Patient Care Technicians — bedside care, mobility support, direct monitoring.
- Licensed Practical/Vocational Nurses (LPNs/LVNs) — in‑person clinical procedures and patient care.
- Electricians — on‑site diagnosis and manual repair of electrical systems.
- Plumbers, Pipefitters, and Steamfitters — complex, variable plumbing and piping work.
- Construction Laborers — general site work that requires physical skills and adaptability.
Why these jobs are the safest under Microsoft’s framework
Microsoft’s analysis finds current LLMs excel at language, research, summarization and routine digital tasks. Jobs that require physical manipulation, on‑site machine operation, complex dexterity or close bodily care score low for AI applicability. In short, blue‑collar roles are less exposed today when work is hands‑on or occurs in unpredictable environments, while certain healthcare roles remain low because they require face‑to‑face care.
“Applicability does not equal job loss — it measures where AI can assist today, not where roles will disappear,” Microsoft emphasizes in its notes.
Important caveats from Microsoft
Microsoft stresses several limits:
- High applicability does not guarantee layoffs, and low applicability does not mean immunity.
- The study focuses on LLM‑type generative AI and excludes robotics, computer vision and other automation that could later affect many fields.
- Even low‑applicability jobs could see AI applied to scheduling, documentation or diagnostics in supporting roles.
Read Microsoft’s further notes on the difference between applicability and displacement: Applicability vs. job displacement.
Implications for Las Vegas, Nevada
Las Vegas’s economy—driven by construction, hospitality and a growing healthcare sector—has local angles tied to Microsoft’s findings. City leaders, employers and training providers can use these insights to align workforce strategy with likely near‑term AI exposure.
- Local labor demand and hiring: The Las Vegas metro continues to need thousands of construction laborers, roofers, plumbers, electricians and machine operators as projects move forward — roles that show low AI applicability and suggest steady in‑person demand. (Context: Full study; GeekWire reporting.)
- Healthcare jobs AI angle: Hospitals, clinics and in‑home care rely heavily on nursing assistants, LPNs and home health aides — roles the study flags as among the safest from LLM disruption. With Nevada’s aging population and tourism workforce needs, training and retention are priorities. (Summary: Fortune.)
- Workforce training and career paths: Community colleges and workforce programs can steer funds toward trades and hands‑on healthcare certificates—short, practical credentials for electricians, HVAC technicians, certified nursing assistants and surgical support roles. (See Microsoft WorkLab materials.)
- Equity and wages: Many low‑applicability roles are lower‑paid today; policymakers should pair upskilling with wage and benefits strategies so these durable jobs provide living incomes.
